Merging time series MAN IP U L ATIN G TIME SE R IE S DATA W ITH - - PowerPoint PPT Presentation

Merging time series

MAN IP U L ATIN G TIME SE R IE S DATA W ITH XTS AN D ZOO IN R

Jerey Ryan

Creator of xts and quantmod

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

Introducing merge()

Combine series by column

cbind() and merge()

Database style joins on index (i.e. by time) Inner, outer, le and right joins

merge(..., fill = NA, join = "outer")

Fill argument handles missingness

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

# Default join = “outer” merge(x, y) x y 2016-08-09 1 2 2016-08-10 1 2 2016-08-11 1 NA 2016-08-12 NA 2 merge(x, y, join = "right", fill = na.locf) x y 2016-08-09 1 2 2016-08-10 1 2 2016-08-12 1 2 merge(x, y, join = "inner") x y 2016-08-09 1 2 2016-08-10 1 2

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

merge(x, c(2, 3, 4)) x c.2..3..4. 2016-08-09 1 2 2016-08-10 1 3 2016-08-11 1 4 merge(x, 3) x X3 2016-08-09 1 3 2016-08-10 1 3 2016-08-11 1 3 merge(x, as.Date(c("2016-08-14"))) x 2016-08-09 1 2016-08-10 1 2016-08-11 1 2016-08-14 NA

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

Introducing rbind()

Combine series by row Rows are inserted in time order All rows in rbind() must have a time The number of columns must match

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

rbind() example

rbind(x, y) x 2016-08-09 1 2016-08-09 2 2016-08-10 1 2016-08-10 2 2016-08-11 1 2016-08-12 2 rbind(x, as.integer(y)) Error in try.xts(c(2L, 2L, 2L)): Error in as.xts.integer( x, ..., .RECLASS = TRUE) :

• rder.by must be either

'names()' or otherwise specified

Let's practice!

MAN IP U L ATIN G TIME SE R IE S DATA W ITH XTS AN D ZOO IN R

Handling missingness

MAN IP U L ATIN G TIME SE R IE S DATA W ITH XTS AN D ZOO IN R

Jerey Ryan

Creator of xts and quantmod

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

Fill NAs with last observation

l.o.c.f. means “last observation carried forward”

na.locf(object, na.rm = TRUE, fromLast = FALSE, maxgap = Inf) cbind(z, na.locf(z), na.locf(z, fromLast = TRUE)) z z.1 z.2 2016-08-09 1 1 1 2016-08-10 NA 1 4 2016-08-11 NA 1 4 2016-08-12 4 4 4

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

Other NA options

Replace NAs na.fill(object, fill, ...) Remove NAs na.trim(object, ...) na.omit(object, ...) Interpolate NAs na.approx(object, ...)

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

NA replace and remove

na.fill(z, fill = -999) z Aug 09, 2016 1 Aug 10, 2016 -999 Aug 11, 2016 -999 Aug 12, 2016 4 Aug 13, 2016 -999 na.trim(z) z Aug 09, 2016 1 Aug 10, 2016 NA Aug 11, 2016 NA Aug 12, 2016 4 na.omit(z) z Aug 09, 2016 1 Aug 12, 2016 4

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

NA interpolation

na.approx() uses index

spacing to linearly approximate the missing values

x x Aug 09, 2016 1 Aug 11, 2016 NA Aug 12, 2016 4 na.approx(x) z Aug 09, 2016 1 Aug 11, 2016 3 Aug 12, 2016 4

Let's practice!

MAN IP U L ATIN G TIME SE R IE S DATA W ITH XTS AN D ZOO IN R

Lags and differences

MAN IP U L ATIN G TIME SE R IE S DATA W ITH XTS AN D ZOO IN R

Jerey Ryan

Creator of xts and quantmod

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

Seasonality and stationarity

Seasonality is a repeating paern Stationarity refers to some bound of the series These paerns are oen compared How get around misalignment of the series?

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

Lagging a time series

Used to align time series for comparisons

lag() will shi observations in time lag(x, k = 1, na.pad = TRUE, ...) k controls number of lags na.pad controls NA introduction

With xts, positive k shis values forward

MANIPULATING TIME SERIES DATA WITH XTS AND ZOO IN R

Differencing series

Convert levels to changes (i.e. deltas)

diff(x,lag = 1, differences = 1, arithmetic = TRUE, log = FALSE, na.pad = TRUE, ...) lag controls which observations arithmetic vs. log calculations

Let's practice!

MAN IP U L ATIN G TIME SE R IE S DATA W ITH XTS AN D ZOO IN R